tocmo0nlord/axolotl
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

use updated version of prebuilt wheels for flash attention for cu130 (#3342)

Browse Source 
* use updated version of prebuilt wheels for flash attention for cu130

* use elif

* fix the uv base installs of FA also

* make wget less verbose
This commit is contained in:
Wing Lian
2026-01-05 13:48:12 -05:00
committed by GitHub
parent b26ba3a5cb
commit 4e61b8aa23
3 changed files with 38 additions and 22 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
docker/Dockerfile-uv-base
View File

@@ -35,8 +35,16 @@ RUN uv pip install packaging setuptools wheel psutil \
&& uv pip install "mamba_ssm @ git+https://github.com/state-spaces/mamba.git@main" \
&& uv pip install awscli pydantic
RUN if [ "$PYTORCH_VERSION" = "2.9.0" ] && [ "$CUDA" = "128" ] ; then \
wget https://github.com/mjun0812/flash-attention-prebuild-wheels/releases/download/v0.4.17/flash_attn-2.8.3+cu128torch2.9-cp311-cp311-linux_x86_64.whl; \
uv pip install --no-cache-dir flash_attn-2.8.3+cu128torch2.9-cp311-cp311-linux_x86_64.whl; \
rm flash_attn-2.8.3+cu128torch2.9-cp311-cp311-linux_x86_64.whl; \
fi
RUN case "$PYTORCH_VERSION" in \
2.9.[0-9]*) \
if [ "$CUDA" = "128" ]; then \
wget -nv https://github.com/mjun0812/flash-attention-prebuild-wheels/releases/download/v0.5.4/flash_attn-2.8.3+cu128torch2.9-cp311-cp311-linux_x86_64.whl; \
uv pip3 install --no-cache-dir flash_attn-2.8.3+cu128torch2.9-cp311-cp311-linux_x86_64.whl; \
rm flash_attn-2.8.3+cu128torch2.9-cp311-cp311-linux_x86_64.whl; \
elif [ "$CUDA" = "130" ]; then \
wget -nv https://github.com/mjun0812/flash-attention-prebuild-wheels/releases/download/v0.5.4/flash_attn-2.8.3+cu130torch2.9-cp311-cp311-linux_x86_64.whl; \
uv pip3 install --no-cache-dir flash_attn-2.8.3+cu130torch2.9-cp311-cp311-linux_x86_64.whl; \
rm flash_attn-2.8.3+cu130torch2.9-cp311-cp311-linux_x86_64.whl; \
fi \
;; \
esac